Linkage analysis of candidate regions using a composite neurocognitive phenotype correlated with schizophrenia

Transcription

1 ORIGINAL RESEARCH ARTICLE (3) 8, & 3 Nature Publishing Group All rights reserved /3 $5. Linkage analysis of candidate regions using a composite neurocognitive phenotype correlated with schizophrenia JF Hallmayer 1,,7, A Jablensky 1,,3, P Michie 1,,6,8, M Woodbury 4, B Salmon, J Combrinck 5, H Wichmann, D Rock, M D Ercole, S Howell, M Dragović and A Kent 1 School of Psychiatry and Clinical Neurosciences, University of Western Australia, Perth, Australia; Centre for Clinical Research in Neuropsychiatry, Graylands Hospital, Perth, Australia; 3 Western Australian Institute of Medical Research, Perth, Australia; 4 Center for Demographic Studies, Duke University, Durham, NC, USA; 5 Community Mental Health Centre, Osborne Park, Perth, Australia; 6 School of Psychology, University of Western Australia, Perth, Australia; 7 Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, CA, USA; 8 School of Behavioural Sciences, University of Newcastle, Newcastle, Australia As schizophrenia is genetically and clinically heterogeneous, systematic investigations are required to determine whether ICD-1 or DSM-IV categorical diagnoses identify a phenotype suitable and sufficient for genetic research, or whether correlated phenotypes incorporating neurocognitive performance and personality traits provide a phenotypic characterisation that accounts better for the underlying variation. We utilised a grade of membership (GoM) model (a mathematical typology developed for studies of complex biological systems) to integrate multiple cognitive and personality measurements into a limited number of composite graded traits (latent pure types) in a sample of 61 nuclear families comprising 8 subjects with ICD-1/ DSM-IV schizophrenia or schizophrenia spectrum disorders and 138 nonpsychotic firstdegree relatives. GoM probability scores, computed for all subjects, allowed individuals to be partly assigned to more than one pure type. Two distinct and contrasting neurocognitive phenotypes, one familial, associated with paranoid schizophrenia, and one sporadic, associated with nonparanoid schizophrenia, accounted for 74% of the affected subjects. Combining clinical diagnosis with GoM scores to stratify the entire sample into liability classes, and using variance component analysis (SOLAR), in addition to parametric and nonparametric multipoint linkage analysis, we explored candidate regions on chromosomes 6, 1 and. The results indicated suggestive linkage for the familial neurocognitive phenotype (multipoint MLS.6 under a low-penetrance model and MLS43. under a high-penetrance model) to a 14 cm area on chromosome 6, including the entire HLA region. Results for chromosomes 1 and were negative. The findings suggest that the familial neurocognitive phenotype may be a pleiotropic expression of genes underlying the susceptibility to paranoid schizophrenia. We conclude that use of composite neurocognitive and personality trait measurements as correlated phenotypes supplementing clinical diagnosis can help stratify the liability to schizophrenia across all members of families prior to linkage, allow the search for susceptibility genes to focus selectively on subsets of families at high genetic risk, and augment considerably the power of genetic analysis. (3) 8, doi:1.138/sj.mp.4173 Keywords: schizophrenia; grade of membership analysis; genetic linkage; variance component analysis; correlated neurocognitive phenotypes; personality traits Introduction Schizophrenia affects approximately 1% of the population worldwide, presenting with similar clinical syndromes and comparable incidence rates across geographic regions. 1 It accounts for.3% of the global burden of disease and shares many of the features of the common complex diseases. In the light of its high Correspondence: Dr J Hallmayer, Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, 11 Welch Road, Room P11, Palo Alto, CA , USA. joachimh@stanford.edu Received 1 June ; revised 13 August ; accepted August heritability, 3 there have been numerous attempts to identify the underlying genetic basis. In the last decade, data from at least 13 genome scans have been published, 4 16 implicating regions on more than half of the chromosomes. 17 None of these findings have yet been consistently replicated, 18,19 although some converging evidence is emerging of linkage on chromosomes 6p, 1p and q. However, the implicated regions are wide and there is considerable disagreement regarding the exact location of the putative genes. Association studies have been similarly inconclusive, notwithstanding some suggestive evidence implicating the serotonin 5HT a and the dopamine DRD 3 receptor genes.

2 51 Both linkage and association research strategies face a common problem in the identification of the relevant phenotype(s). The majority of genetic studies remain predicated on a clinical symptom-based phenotype, defined by the current diagnostic criteria of the disorder. However, finding individual susceptibility loci using diagnosis as a phenotype may be problematic. 1 3 Whereas schizophrenia is broadly heritable as a complex clinical entity, its symptoms show significant inter- and intra-individual variation, both cross-sectionally and over time, 4 and may not provide robust phenotypes for genetic study. Moreover, the genotypes underlying schizophrenia tend to remain clinically unexpressed in the majority (8 9%) of first-degree relatives. 5 The requirement for a phenotype which is stable and has a high relative risk ratio, 6 l R, can be addressed by augmenting the diagnostic description of schizophrenia with measurements of heritable biological traits that are plausibly implicated in pathogenesis and may be more directly influenced by genes coding for structural proteins or signalling molecules than clinical symptoms, such as delusions or hallucinations. Recent research has revealed a range of neurocognitive, neurophysiological and neuroanatomic abnormalities in nonpsychotic biological relatives of probands with schizophrenia, which possess many of the characteristics of vulnerability markers and are relatively specific to schizophrenia as compared to other psychotic disorders. 7 3 Terms such as endophenotypes, correlated phenotypes, or level II indicators are used to describe this intermediate level. A genetic relationship between these phenotypes and schizophrenia is strongly suggested by studies on monozygotic twins discordant for the clinical disorder. 31,3 The notion that using such correlated traits might dramatically enhance the capacity to detect linkage is supported by recent findings. Thus, Arolt et al 33 scanned the 6p1 3 region with 16 microsatellite markers to test for linkage to eye tracking dysfunction and schizophrenia. Two-point lod scores of 3.51 and 3.44 were found for markers D6S71 and D6S8 which, when included in a multipoint analysis, yielded a lod score of 4.. Using both parametric and nonparametric (sib-pair) analysis of 14 members from nine families phenotyped for the auditory evoked potential P 5 (indexing a sensory gating defect in schizophrenia patients), Freedman et al 34 reported a multipoint lod score of 5.9 for a marker on chromosome 15q13 14, which falls within the region coding for the a7- nicotinic cholinergic receptor. A genome-wide linkage analysis, screening for loci underlying a composite inhibitory phenotype combining the P 5 event-related potential and antisaccade oculomotor performance measures, 35 produced evidence for linkage (lod score 3.55) to the DS315 marker on chromosome q. Further promising endophenotypes for schizophrenia include reduced hippocampal volume 36 or neurocognitive deficits in attention control and executive function. 37 Faraone et al 38 estimated familial risk ratios for several phenotypes correlated with schizophrenia and concluded that among all the measures examined, an attentional deviance index, derived by Cornblatt et al 39 and based mainly on the identical pairs version of the Continuous Performance Task (CPT-IP), yielded the highest familial risk ratio of 3. Analysing differences in the profiles of neuropsychological deficits in twins discordant for schizophrenia, Cannon et al 3 suggested that at least partially distinct sets of genes for schizophrenia contribute to such variation, and concluded that measurement of neurocognitive profiles in patients and first-degree relatives may enable the determination of the segregation of specific deficits within families. We report results from a family study in which we constructed a set of neurocognitive phenotypes correlated with schizophrenia by integrating measurements on multiple neurocognitive tasks and personality traits acquired from patients with schizophrenia and schizophrenia spectrum disorders, their biological first-degree relatives, and nonpsychiatric control subjects. We used grade of membership (GoM) analysis, a statistical technique based on a fuzzy sets mathematical model, 4,41 to derive an empirical taxonomy of neurocognitive deficits and behavioural traits in schizophrenia that excluded clinical variables. External criteria, including clinical symptoms and other descriptors of illness, were later adduced to test the construct and predictive validity of the resultant taxonomy of neurocognitive deficit as putative endophenotypes. We describe firstly the results of variance component linkage analysis between markers in three candidate regions (chromosomes 6, 1 and ) and the neurocognitive endophenotypes. Secondly, we report on parametric linkage analysis, carried out by combining the clinical phenotype with the neurocognitive endophenotypes. Thirdly, we present additional quantitative trait analyses for selected correlated phenotypes. To our knowledge, this is the first application of GoM to the genetic analysis of neurocognitive deficits in schizophrenia. Materials and methods Family ascertainment Families were recruited by screening consecutive admissions to a major psychiatric hospital. Inclusion criteria were: (i) probands (age 15 54) met both ICD- 1 and DSM-IV criteria for a lifetime diagnosis of schizophrenia or schizophrenia spectrum disorder; and (ii) at least one sibling and one parent were available for the study. Schizophrenia spectrum diagnoses included schizoaffective disorder, schizotypal and paranoid personality disorder, and other nonaffective psychoses (excluding persistent delusional disorders). Families were excluded if the proband had a comorbid organic brain disease or substance-use disorder that could account for the psychotic symptoms, or had language difficulties

3 likely to interfere with interviewing or testing. The study protocol was explained to all participants and written informed consent was obtained. The Committee for Human Rights of the University of Western Australia and the Graylands Hospital Ethics Committee (Perth, Western Australia, Australia) approved the study. Clinical evaluation Probands and relatives were interviewed using the Schedule for Clinical Assessment in Neuropsychiatry, SCAN, Version.. 4 Full pedigree description and history of psychiatric and general morbidity were obtained from a key family member, using the NIMH Family Interview for Genetic Studies, FIGS. 43 Consensus diagnostic evaluation was carried out by two senior clinicians reviewing independently all diagnostic information and assigning ICD-1 Diagnostic Criteria for Research (DCR) 44 and DSM-IV 45 lifetime diagnoses, being blinded to family relatedness and test scores of the subjects. All diagnostic disagreements were resolved through a joint case review. Neurocognitive assessment protocol Probands, relatives and controls completed a neurocognitive battery assessing seven domains of function. General intelligence: The National Adult Reading Test, NART 46 provided an estimate of premorbid WAIS-R IQ; the Shipley Institute of Living Scale 47 was used to estimate current intellectual functioning, converted to WAIS-R IQ. Sustained attention: Two high-processing load versions of the Continuous Performance Task (CPT) were used: the degraded stimulus version, CPT-DS, 48 and the identical pairs version, CPT-IP. 49 Whereas CPT-DS involves an increased demand on visual encoding, CPT- IP selectively challenges working memory. Hit rate, reaction time for hits, error of commission rate (false alarms), and overall measures of signal detection, dl, and bias, cl 5 were derived for both tasks. Verbal fluency: The FAS version of the Controlled Oral Word Association Task, 51 a measure of effortful lexical retrieval, was administered to tap a component of executive function. Verbal memory: The Rey Auditory Verbal Learning Test, RAVLT, 5 was used to measure immediate and delayed recall of lists of words, shortterm retention following distraction, and errors (irrelevant intrusions of nonlist words). Speed of processing: The Inspection Time, IT, task 53 provided a measure of perceptual encoding and speed of processing. The test measures the minimum duration of stimulus presentation calibrated to achieve 7.7% correct responses on a visual discrimination task and uses a superimposed backward mask to manipulate the duration of stimulus exposure. Handedness and motor speed lateralisation: The Edinburgh Inventory 54 of hand/foot/eye preference for 1 activities and the Finger Tapping Task 55 measuring tapping speed and inter-tap interval variability for each hand were administered to all participants. Personality trait measures All subjects completed the Temperament and Character Inventory, TCI, 56 and the Schizotypal Personality Questionnaire, SPQ. 57 GoM analysis The neurocognitive and personality trait data were analysed using the GoM methodology which integrates multiple measurements into a limited number of composite graded traits. GoM, a mathematical typology developed for studies of complex biological systems, 4,41 is one of the family of latent structure analysis methods 58 and is based on the concept of fuzzy sets. It resolves heterogeneity across multiple attributes (eg symptoms, psychological test measurements, or any other biological variable) by assigning each individual to several analytically derived latent classification categories or pure types (PTs) and quantifies the degree to which an individual belongs to any given PT. PTs are described by profiles of attributes where the probability (l kj ) of each attribute is estimated, being manifested by an individual belonging entirely to one PT. PT coefficients (l kj ) and scores (g ik ) on individual membership in the PTs are estimated simultaneously from the data matrix by maximum likelihood. The g ik scores of an individual on all PTs are constrained to be non-negative and sum up to 1., thus enabling a quantification of an individual s membership in more than one PT. GoM has been used to analyse large data sets from the WHO International Pilot Study of Schizophrenia 59 and a historical cohort of E Kraepelin s dementia praecox cases; 6 it has also been used in genetic research into Alzheimer s disease; 61 and in a replication of mapping the IDDM11 locus in diabetes. 6 Using GoM and 68 input variables, we obtained an optimal partitioning of the data set into five PTs. As a concurrent validator, we performed latent class analysis of the same data set, which also resulted in five classes with a 6% overlap (P ¼.) in the clustering of the variables by the two methods. Each PT identifies a particular idealised multivariate configuration of neurocognitive and personality traits, approximated by the real individuals in the sample to varying degrees, specified by their GoM scores (g ik ) on each one of the PTs. Assignment of principal membership to one of the PTs is based on the observed greatest deviation of each individual s scores from the mean g ik for each PT. Thus, the study population was distributed into five clusters, with individuals within each cluster approximating the latent structure of one of the PTs. PT I: Represented by probands with schizophrenia and a small number of affected siblings; characterised by both low premorbid and low current IQ; high rate of irrelevant intrusions on delayed word recall; abnormal performance on both sustained attention tasks, with a particularly marked deficit on the degraded stimulus task (CPT-DS), suggesting a predominantly frontoparietal impairment. This PT also exhibits more nonright handedness and soft 513

4 514 neurological signs than any other GoM type. It is further characterised by high scores on positive schizotypy (SPQ odd beliefs, magical ideation) and on TCI novelty seeking. PT II: Includes probands as well as unaffected siblings and parents. Both premorbid and current IQ are low, but in contrast to PT I, there is a significant postonset drop in IQ from the premorbid to the current level. Compared to PT I, this type is more impaired on verbal fluency and on CPT-IP, the attention task challenging working memory, and less impaired on the degraded stimulus task CPT-DS, which assesses attentional control of visual encoding. It exhibits a high score on negative schizotypy (SPQ constricted affect) and lower scores than PT I on odd beliefs and magical ideation. PT III: The profile is normative, in the sense that performance across all tasks is within 71SD of the control means. It includes mainly unaffected relatives and controls, as well as a minority of schizophrenia patients with apparently less impaired neurocognitive performance. PT IV: Expressed in unaffected relatives and controls, includes a small number of probands. It differs from PT III by a lower premorbid IQ and a higher rate of nonright handedness. PT V: Consists mainly of parents, controls, and a small number of probands with schizophrenia over age 55. It is characterised by a relatively high premorbid IQ and an age-related decline of general ability. With the exception of a low positive correlation (r ¼.13) between PT I and PT II, the PTs were uncorrelated. Within each cluster of subjects assigned to the same PTs, the neurocognitive profiles were qualitatively similar for clinically affected and unaffected individuals. However, as shown in PT II (Figure 1), schizophrenia patients display more conspicuous deviations than their unaffected relatives on measures of general ability, sustained attention and verbal memory. Clinical correlates of GoM PTs We selected a priori 17 SCAN symptom clusters indexing positive symptoms, negative symptoms, affective symptoms, and disorganisation. Comparison of the symptom profiles indicated that whereas both PT I and PT II cases scored high on delusions and auditory hallucinations, PT I subjects had higher scores on elation, Schneiderian first-rank symptoms, intensity of delusions, bizarre delusions and positive thought disorder. PT II subjects, on the other hand, had higher scores on negative thought disorder (poverty of content of speech, restricted quantity of speech, nonsocial speech) and flat or incongruous affect (blunting, incongruity, reduced facial and voice expression). The comparison of the research diagnoses of ICD-1 clinical subtypes of schizophrenia and related disorders revealed a significant difference (P ¼.15) in the ratios of paranoid to nonparanoid schizophrenia within PT I and PT II. While paranoid schizophrenia was the modal (39.5%) diagnosis of subjects assigned to PT II, it accounted for only 8.9% of the diagnoses in PT I (the odds ratio for a subject with a diagnosis of paranoid schizophrenia to be PT II, rather than PT I, was 5.35; CI ). In contrast, nonparanoid subtypes were over-represented (55.3%) in PT I. Notably, all cases diagnosed with rare or atypical forms of schizophrenia (catatonic, hebephrenic, simple, other and unspecified ) turned out to have the PT I neurocognitive phenotype. Furthermore, PT I patients had a greater mean number 4 3 z-scores baseline = performance of controls Affected (N=43) Unaffected (N=1) -5 Premorbid IQ Current IQ Verbal fluency RAVLT immediate recall RAVLT delayed recall RAVLT delayed intrusions CPT-IP/dL CPT-DS/dL Laterality quotient TCI novelty seeking TCI harm avoidance TCI self-transcendence SPQ odd beliefs SPQ odd speech SPQ constricted affect SPQ total score Figure 1 Neurocognitive profiles of affected probands and unaffected first-degree relatives assigned to PT II.

5 of hospital admissions (8.8) than patients in any of the other PTs, as well as a greater mean duration of hospitalisations (15.1 weeks, compared to 11.7 weeks for PT II and 9.6 weeks for PTs III V combined). The prescribed dosages of typical and atypical antipsychotic drugs at the time of assessment differed across patient groups assigned to different PTs. Adding up the chlorpromazine-equivalent doses, PT I patients had the highest prescribed mean dose (747 mg/day), compared to 57 mg/day for PT II patients and 38 mg/day for PTs III V patients. It appeared, therefore, that the GoM neurocognitive phenotypes, especially PT I and PT II, were correlated with distinct clinical profiles of schizophrenia. Familial aggregation of GoM PTs Using ANOVA, we found a highly significant (Po.) familial aggregation of GoM scores for PT II, significant for PT III (Po.4) and PT IV (Po.3), but no significant familial aggregation for PT I and PT V (Table 1). The additive genetic heritability (using maximum likelihood variance decomposition as implemented in SOLAR 63 ) was.4 for PT II and.35 for the combined PT I and PT II phenotype (Table 1). A plausible interpretation of these data is that PT I and PT II, which select the majority of probands with schizophrenia, represent multivariate phenotypes differing in the strength of their association with the genetic liability to schizophrenia. Cell lines and genomic DNA Permanent cell lines were established from all probands by transforming lymphocytes with Epstein Barr virus using standard techniques. 64 Genomic DNA was prepared from whole blood or cell lines using the salting out method. 65 Polymerase chain reaction (PCR) assays were performed in 8 ml volume reactions containing 1 ng genomic DNA;.5 mm of each dntp;..6 mm of each primer pair labelled with Fam, Hex, or Tet;.8 ml of 1 buffer (Perkin-Elmer, Norwalk, CT, USA); mm MgCl ; and. units AmpliTaq Gold Polymerase (Perkin Elmer). PCR assays were performed in an Applied Biosystems (Foster City, CA, USA) 96 Thermocycler by denaturing for 1 min at 941C, then 15 cycles of 3 s at 941C, 15 s at 61C, and 15 s at 71C. Another cycles of 3 s at 941C, 15 s at 651C, and 15 s Table 1 Heritability estimates a (Hr) for GoM PTs 1 5 GoM PT Hr Standard error P PT I. F.5 PT II PT III PTI V PT V. F.5 Combined PT I and PT II a Calculated using the variance-component methodology implemented in SOLAR. 63 at 71C were performed, followed by a final extension step of 1 min at 71C. Fluorescently labelled markers were analysed on Applied Biosystems 373 or 31 Genetic Analysers, and the data were further analysed using Applied Biosystems GeneScan and Genotyper software. Polymorphic bands were scored and alleles were assigned to the pedigree members. Markers typed Chromosome 6: D6S63; D6S49; D6S85; D6S19; D6S76; D6S114; D6S439; D6S497; D6S119; D6S46; D6S57; D6S445; D6S1613; D6S44; D6S83; D6S31; D6S68; D6S474; D6S67. Chromosome 1: D1S191; D1S143; D1S197; D1S146; D1S64. Chromosome : DS4; DS539; DS315; DS8; DS78; DS83; DS43; DS74; DS1169. Linkage analyses Parametric and nonparametric multipoint analyses were carried out using GENEHUNTER. 66 Quantitative trait analyses were performed with the SOLAR program. 63 Allele frequencies for all calculations were derived from unrelated individuals. We used a combination of clinical diagnosis and GoM scores (g ik ) to stratify the entire sample (ie, both clinically affected and clinically unaffected individuals) into 1 liability classes (Table ). For parametric linkage analyses, all individuals with a diagnosis of schizophrenia or schizophrenia spectrum disorder were assigned to liability class I. Liability class assignment for all remaining, clinically unaffected subjects was determined by their GoM scores (g ik )on the PT II composite neurocognitive phenotype which displayed the highest degree of familial aggregation. As the lack of significant heritability for PT I does not preclude influence of genetic factors on its expression, combined PT I and PT II scores were used to stratify the sample in an alternative analysis. Two genetic models were calculated: a low-penetrance model with a maximum penetrance of., and a high-penetrance model with a maximum penetrance of.9. Disease frequency was set at.1 and., respectively. Separate analyses were carried out with those single components of the composite phenotype which made the greatest contribution to the maximum likelihood function differentiating the GoM PTs: general ability (premorbid IQ, current IQ); sustained attention (CPT-IP); and delayed word recall (RAVLT). Results A total of 61 families (comprising 59 individuals) with at least one proband diagnosed with schizophrenia or schizophrenia spectrum disorder were potentially available for linkage analyses (we did not extend the families beyond first-degree relatives). Among the family members, 1 were deceased; 14 refused to participate in the study; and another five 515

6 516 Table Liability classes assigned in the parametric linkage analysis PT II scores (g ik ) Liability class Low-penetrance model High-penetrance model DD Dd dd DD Dd dd I II III IV V VI VII VIII IX X Linkage was calculated for two different genetic models, one with a maximum penetrance value of. (low-penetrance model) and a gene frequency for the disease allele.1, and the other with a maximum penetrance of.9 and a gene frequency for the disease allele.. All individuals with a diagnosis of a schizophrenia or schizophrenia spectrum disorder were assigned to liability class I. For all other family members, PT II or PT I and PT II combined g ik scores were used for liability class assignment. could not be diagnostically evaluated. Besides the 61 probands (48 with narrow schizophrenia and 13 with schizophrenia spectrum diagnoses), an additional 19 first-degree relatives received a diagnosis of either schizophrenia (five subjects) or schizophrenia spectrum disorder (14 subjects). Another 64 relatives received lifetime diagnoses of various nonpsychotic affective, anxiety or personality disorders; and 84 relatives were assessed as having no psychiatric disorder. The 148 diagnostically evaluated relatives, unaffected with schizophrenia or schizophrenia spectrum disorders, contributed information to linkage analyses based on the GoM scores of their neurocognitive phenotypes. Excluding the five relatives who declined a diagnostic interview, genotypes from a total of 8 individuals were used for linkage analyses. Variance component linkage analysis We first performed two-point linkage analysis using the variance component procedure implemented in SOLAR. None of the markers typed on chromosome 1 or showed any lod scores above.7. A maximum two-point lod score of.65 was obtained on chromosome 6 with marker D6S76 for the combined PT I and PT II phenotype. Six more markers on chromosome 6 resulted in lod scores above 1. (D6S114, D6S1613, D6S49, D6S57, D6S46, D6S119). The maximum lod score for PT II (1.43) was obtained with the same marker D6S76 (Table 3). In a second step, we carried out multipoint variance component linkage analysis at constant increments of 1 cm (Figure 1). The maximum lod score observed for combined PT I and PT II was.51 at 44 cm. For PT II, the maximum lod score was 1.59 at 56 cm. As with the two-point analyses, none of the lod scores were above.6 on chromosomes 1 and (Figure ). Variance component linkage analysis between the same markers and single neurocognitive measures used as correlated phenotypes (CPT-IP, current IQ, premorbid NART IQ, and RAVLT total word recall) resulted in only weakly positive lod scores (Table 4). Parametric linkage analysis Results from the parametric linkage analysis are shown in Figure 3. Two peaks about 4 cm apart were observed for both PT II and the combined PT I/PT II phenotype. The highest peak with a maximum lod score of.11 was calculated at D6S439 for the combined phenotype using the low-penetrance model. However, the peak for the high-penetrance model was only marginally lower (.) as were the peaks for the PT II phenotype (MLS.1 for the low-penetrance model and 1.98 for the high-penetrance model). A second, lower peak was calculated close to the marker D6S76. In order to test for genetic differences between the two phenotypes (PT I and PT II) we subdivided the families on the basis of the ratio of the GoM scores (g ik ) PT I/PT II, for each proband. PT I probands were defined by a PT I/ PT II ratio of greater than 1.5 and PT II probands by a PT I/ PT II ratio of less than 1.5. For families with a PT II proband, we obtained maximum lod scores above.6 at D6S76 and D6S439 (Figure 4). s for PT I families at the same locations were negative ( 1.16 and.68, respectively). The highest positive lod score for the PT I subgroup was only slightly above 1 between the markers D6S474 and D6S76, 7 cm away from the peak of the families with PT II probands. In a final analysis, we increased penetrance values for liability class 1 to.9, setting the other liability class values accordingly (see Table ). The frequency of the disease allele was assumed to be only.. Under this model, the maximum lod scores were above 3 for the PT II subgroup of families.

7 Table 3 Two-point lod scores for variance component linkage analysis 517 Marker PT II a PT I and PT II b Marker PT II a PT I and PT II b Chromosome 6 Chromosome 1 D6S D1S D6S D1S197.. D6S D1S D6S19.. D1S64.. D6S D1S D6S D6S Chromosome D6S D6S DS4.8.4 D6S DS539.. D6S DS315.. D6S DS8..4 D6S DS78.. D6S44.. DS83.. D6S83.. DS43.. D6S DS74.. D6S68.. DS D6S D6S67..1 a Composite neurocognitive phenotype, GoM PT II. b Composite neurocognitive phenotype, GoM PTs I and II combined Chromosome 6p Chromosome 1p Chromosome 4 6 Figure Results from multipoint variance component linkage analysis for chromosome 6p, 1p, and markers. The solid lines plot the lod scores for PT II and the dotted lines for the combined PTs I and II phenotype. Table 4 Maximum lod scores for variance component linkage analysis with selected single neurocognitive measures as correlated phenotypes Phenotype Discussion Chromosome 6 Chromosome 1 Chromosome CPT-IP a Current IQ b.68.. Premorbid IQ (NART c ) RAVLT d delayed word recall a Continuous Performance Task, Identical Pairs. b Shipley Institute of Living Scale, converted to estimated WAIS-R IQ. c National Adult Reading Test. d Rey Auditory Verbal Learning Test. Despite the high heritability of schizophrenia, identifying susceptibility genes contributing to its aetiology has proven to be a daunting task. 67 Schizophrenia is a complex disorder 68 characterised by a widely varied clinical phenotype, low penetrance and existence of phenocopies. The genetic model currently commanding broad agreement is one of multiple genes of small-to-moderate effect increasing the risk of

8 Chromosome 6p Chromosome 1p Chromosome 6p Chromosome 1p Chromosome Chromosome Figure 3 Results from parametric multipoint linkage analysis for markers on chromosomes 6p, 1p and. Individuals with a diagnosis of schizophrenia or a schizophrenia spectrum disorder were considered affected. Unaffected relatives contributed information to linkage analyses based on their GoM scores. Left panel: graphs based on PT II scores; right panel: graphs based on combined PTs I and II scores. Dotted lines: high-penetrance model; solid lines: low-penetrance model. 4 Chromosome 6p Figure 4 Results from parametric multipoint linkage analysis for chromosome 6p. Individuals with a diagnosis of schizophrenia or a schizophrenia spectrum disorder were considered affected. Unaffected relatives contributed information to linkage analyses based on their PT II scores. Families were subdivided based on the ratio of PT I to PT II scores. Open circles: multipoint lod scores for the low-penetrance model and families with a PT I/PT II score ratio greater than 1.5. Solid squares: low-penetrance model and families with a PT I/PT II score ratio less than 1.5. Open triangles: high-penetrance model and families with a PT I/PT II score ratio less than 1.5.

9 schizophrenia through epistatic gene gene interactions, as well as interactions with nongenetic risk factors. 69 Under this model, the power to detect susceptibility genes depends strongly on the magnitude of the effect of the individual gene under investigation. 7 For an oligogenic disorder, the sample size required is very large 71 and the difficulty in recruiting a sufficient number of families is a major constraint. 8 In the case of schizophrenia, the majority of probands do not have a first-degree relative affected with the same disorder, and therefore only a minority of potential families are informative for linkage analysis. 7,73 Thus, studies based exclusively on diagnosis are unlikely to generate adequate power to detect susceptibility genes. In the present study, we applied a novel approach, capable of overcoming some of the limitations of clinical diagnosis as phenotype, by identifying composite quantitative traits indexing neurocognitive deficits known to occur both in patients with schizophrenia and in a proportion of their unaffected first-degree relatives. We used a mathematical typology based on fuzzy statistical models, 4,41 which identifies latent patterns, or PTs, among multiple variables and allows individuals to be partly assigned to more than one such PT by computing GoM scores for each subject on all of the PTs. Using as input variables 68 neurocognitive and personality test measures, but no symptoms or clinical history items, we applied a maximum likelihood criterion to obtain an optimal partitioning of the data matrix into five PTs, each identifying a unique profile of neurocognitive and personality traits. Two of the PTs (PT I and PT II) were confined to probands with schizophrenia and their first-degree relatives, whereas PTs III V were mainly represented by unaffected relatives and controls. PT I and PT II were associated with independently assessed, distinct clinical profiles of lifetime symptoms and severity of illness. Whereas PT II, and to a lesser degree PTs III and IV, aggregated significantly in families, PTs I and V did not. However, the combination PT I and PT II did show significant familiality. As it accounted for 74% of the schizophrenia patients in the sample, we assumed that the conflation of these two PTs represents a correlated phenotype that reflects genetic vulnerability to schizophrenia. Compared to conventional linkage analysis, the correlated phenotype approach, as applied here, offers three advantages. Firstly, it allows inclusion of families ascertained through a proband with schizophrenia, irrespective of the presence or absence of other clinically affected family members. Secondly, as PT coefficients (g kj ) treat susceptibility to neurocognitive dysfunction as a quantitative trait (instead of a dichotomy of affected and unaffected status), all individuals in the study sample are potentially informative. Thirdly, by comparing the analysis of a dichotomised trait with that of a quantitative version of the same trait, it can be shown that both precision and power to detect linkage increase when the trait is coded as a continuous variable. 74,75 If a composite trait is constructed, integrating a correlated structure of multiple traits, power can be further increased. 76,77 Based on published results, 9,14,78 83 we selected markers on chromosomes 6, 1 and, and tested them for linkage. Variance component linkage analysis revealed maximum positive lod scores above.5 for both two-point and multipoint analysis with markers on chromosome 6 for the combined PT I and PT II phenotype (lod scores for PT II alone were also positive but overall lower). The peak lod scores for the combined PT I and PT II phenotype were calculated with marker D6S76, located at the telomeric end of the HLA region. In a second set of calculations, we included the ICD-1/DSM-IV diagnosis of schizophrenia in the definition of the affection status but retained in the analysis all unaffected individuals by stratifying them into liability classes on the basis of their PT II or PT I and PT II g ik scores. Using this approach enabled an increase in the power to detect linkage since not only affected but also a large proportion of the clinically unaffected individuals became informative. Power calculations based on the 61 families using the SLINK program demonstrated a nearly six-fold increase in the average lod score for a linked marker to a putative disease locus (Table 5). This approach is not restricted to families with more than one family member affected with schizophrenia and therefore allows aggregating much larger data sets. As with variance component analysis, the highest positive lod scores were calculated with markers from chromosome 6. The peak lod scores of about were closer to the centromeric end of the HLA region at D6S439. The low heritability of PT I, together with the finding of mixed handedness and soft neurological signs, argues in favour of a nongenetic contribution to the aetiology of the schizophrenic syndrome in this subset of families. We therefore subdivided the families on the basis of the probands PT I and PT II g ik scores. As the g ik scores of an individual on all PTs are constrained to be non-negative and sum up to 1., they enable a quantification of an individual s membership in more than one PT. We considered a family to be PT I if the PT I score of the proband was at least 1.5 times the PT II score. A total of 1 families (34.4% of all families) with a proband having a ratio of PT I to PT II scores 1.5 were identified and analyses were conducted for both groups of families separately (Figure 4). Multipoint lod scores for the two groups were strikingly different. s for PT I families were negative for markers that resulted in positive results in PT II families and positive for markers that resulted in negative lod scores in PT II families. For the PT II sample, lod scores above. were calculated, stretching over a 14 cm area including the entire HLA region. s greater than.6 in the low-penetrance model increased to above 3. in the high-penetrance model. 519

10 5 Table 5 Simulated average and maximum lod scores for 61 families used in the linkage study Low-penetrance model High-penetrance model Sz only a Sz+PT II b Sz+PT I/PT II c Sz only a Sz and PT II b Sz+PT I/PT II c Average Maximum a Affected subjects with schizophrenia or schizophrenia spectrum disorders only. b Affected subjects with schizophrenia or schizophrenia spectrum disorders, plus unaffected relatives, liability stratified by GoM scores (g ik ) for PT II. c Affected subjects with schizophrenia or schizophrenia spectrum disorders, plus unaffected relatives, liability stratified by GoM scores (g ik ) for PTs I and II combined. Genetic models and liability classes assigned as described in Table and text. Results are based on 1 simulations using SLINK (Ott, 1989; Weeks et al, 199), assuming a marker with heterozygosity of.7 located 5 cm distant to a putative disease locus. One possible explanation of the results is that PT II may be constrained by a different genetic profile than PT I, suggesting that different components of the schizophrenia phenotype are influenced by different genetic loci. As it is unlikely that any of the genes involved code directly for schizophrenic signs and symptoms, such as delusions, hallucinations or thought disorder, we assume that the putative genes influence their occurrence indirectly via combinations and interactions of basic processes at synaptic and neural circuitry level, phenotypically expressed as neurocognitive deficits and personality traits. Notably, variance component linkage analysis of single neurocognitive measures, such as the CPT-IP, which contribute to the differentiation of PT I and PT II, did not show any lod scores greater than.7, and positive lod scores above. were restricted to the multivariate phenotype. Therefore, the genetic effects are more likely to involve multiple intermediate and interactive steps, thus resulting in a very complex relationship between genes and phenotype. We suggest that the multivariate PT II neurocognitive phenotype can be viewed as a pleiotropic manifestation of genes underlying the susceptibility to a clinical form of schizophrenia characterised primarily by paranoid delusions and negative symptoms. The results presented provide suggestive evidence 84 of linkage for susceptibility gene(s) for schizophrenia located on the short arm of chromosome 6. Several other groups have reported positive linkage findings on chromosome 6p, 4,6,1,14,78,8,81,85 87 and it has been argued that the 6p findings reach the threshold 84 for significant linkage. However, the locations of the markers showing the strongest evidence for linkage vary substantially across the studies. In this present study, maximum positive lod scores were calculated with markers close to or within the HLA region. One possibility is that HLA antigens themselves confer susceptibility to schizophrenia. 88 Alternatively, this effect may be caused by another gene in close proximity. 89 The HLA region is extremely gene rich (the HLA class III region alone contains at least 59 genes and is the highest gene density region. 9 Positive associations have been reported between schizophrenia and polymorphisms in the NOTCH4 gene. 91 However, replication attempts have been largely negative A potential role of the tumour necrosis factor alpha (TNFa) gene for susceptibility to schizophrenia has been proposed by Bouin et al, 97 who reported a positive association between a biallelic base exchange polymorphism that directly affects TNFa plasma levels and schizophrenia. TNFa, produced by glial cells, influences synaptic plasticity and modulates responses to neural injury. 98 More recently, a significant association in the region 6p4 1 has been reported for the dysbindin gene. 99 Considering the complex inheritance and likely genetic heterogeneity of schizophrenia, it may be extremely difficult, if not impossible, to demonstrate an aetiological role for any of these genes. However, by identifying susceptible individuals based on indicators that are more proximal to the genetic liability, it may be possible to obtain more coherent genetic findings. The results reported here indicate that neurocognitive and personality measures in schizophrenia patients and their biological relatives aggregate nonrandomly into distinct patterns or PTs. Two of the pure types (PT I and PT II) select the majority of clinically affected subjects and thus index a high risk for schizophrenia. Of these, PT II is familial and heritable, whereas PT I suggests a greater influence of environmental factors in its aetiology. The distinction between these two types in the genetic analysis of schizophrenia may be a step towards identifying more homogeneous samples for both linkage and association studies. Acknowledgements This work was supported by the National Health and Medical Research Council (NHMRC), Australia (grants 96579, 9971 and 196) and the Graylands Hospital (Perth, Western Australia, Australia). The following individuals made specific contributions to study design, data collection, data analysis, and manuscript editing. Perth: J Box, JE Cooper, B

11 Jansen, J Johnston, R Kaiser, V Morgan, R Stienstra, J Todd, D Vile, D Wood, and P Wynn Owen; Duke University, Durham, NC, USA: E Corder, J Wagner, K Manton. The cooperation of the clinical and administrative staff at Graylands Hospital, Perth, is gratefully acknowledged. We especially thank the patients, family members, and other volunteers who participated in this study. References 1 Jablensky A, Sartorius N, Ernberg G, Anker M, Korten A, Cooper JE et al. Schizophrenia: manifestations, incidence and course in different cultures. A World Health Organization Ten-Country Study. Psychol Med Monogr Suppl 199; : Murray CJL, Lopez AD. The Global Burden of Disease. A Comprehensive Assessment of Mortality and Disability from Diseases, Injuries, and Risk Factors in 199 and Projected to. The Harvard School of Public Health, distributed by Harvard University Press: Cambridge, MA, Kendler KS, Diehl SR. The genetics of schizophrenia: a current, genetic epidemiological perspective. Schizophr Bull 1993; 19: Bailer U, Leisch F, Meszaros K, Lenzinger E, Willinger U, Strobl R et al. Genome scan for susceptibility loci for schizophrenia. Neuropsychobiology ; 4: Blouin JL, Dombroski BA, Nath SK, Lasseter VK, Wolyniec PS, Nestadt G et al. Schizophrenia susceptibility loci on chromosomes 13q3 and 8p1. Nat Genet 1998; : Brzustowicz LM, Hodgkinson KA, Chow EW, Honer WG, Bassett AS. Location of a major susceptibility locus for familial schizophrenia on chromosome 1q1 q. Science ; 88: Coon H, Jensen S, Holik J, Hoff M, Myles-Worsley M, Reimherr F et al. Genomic scan for genes predisposing to schizophrenia. Am J Med Genet 1994; 54: DeLisi LE, Shaw SH, Crow TJ, Shields G, Smith AB, Larach VW et al. A genome-wide scan for linkage to chromosomal regions in 38 sibling pairs with schizophrenia or schizoaffective disorder. Am J Psychiatry ; 159: Faraone SV, Matise T, Svrakic D, Pepple J, Malspina D, Suarez B et al. Genome scan of European-American schizophrenia pedigrees: results of the NIMH Genetics Initiative and Millennium Consortium. Am J Med Genet 1998; 81: Hovatta I, Varilo T, Suvisaari J, Terwilliger JD, Ollikainen V, Arajarvi R et al. A Genome-wide screen for schizophrenia genes in an isolated Finnish subpopulation, suggesting multiple susceptibility loci. Am J Hum Genet 1999; 65: Levinson DF, Mahtani MM, Nancarrow DJ, Brown DM, Kruglyak L, Kirby A et al. Genome scan of schizophrenia. Am J Psychiatry 1998; 155: Lindholm E, Ekholm B, Shaw S, Jalonen P, Johansson G, Pettersson U et al. A schizophrenia-susceptibility locus at 6q5, in one of the world s largest reported pedigrees. Am J Hum Genet 1; 69: Pulver AE, Lasseter VK, Kasch L, Wolyniec P, Nestadt G, Blouin JL et al. Schizophrenia: a genome scan targets chromosomes 3p and 8p as potential sites of susceptibility genes. Am J Med Genet 1995; 19: Schwab SG, Hallmayer J, Albus M, Lerer B, Eckstein GN, Borrmann M et al. A genome-wide autosomal screen for schizophrenia susceptibility loci in 71 families with affected siblings: support for loci on chromosome 1p and 6. Mol Psychiatry ; 5: Shaw SH, Kelly M, Smith AB, Shields G, Hopkins PJ, Loftus J et al. A genome-wide search for schizophrenia susceptibility genes. Am J Med Genet 1998; 81: Williams NM, Rees MI, Holmans P, Norton N, Cardno AG, Jones LA et al. A two-stage genome scan for schizophrenia susceptibility genes in 196 affected sibling pairs. Hum Mol Genet 1999; 8: Pulver A. Search for schizophrenia susceptibility genes. Biol Psychiatry ; 47: DeLisi L, Crow T. Chromosome Workshops 1998: current state of psychiatric linkage. Am J Med Genet 1999; 88: Badner JA, Gershon ES. Meta-analysis of whole-genome linkage scans of bipolar disorder and schizophrenia. Mol Psychiatry ; 7: O Dononvan MC, Owen MJ. Candidate-gene association studies of schizophrenia. Am J Hum Genet 1999; 65: Ginsburg BE, Werick TM, Escobar JI, Kugelmass S, Treanor JJ, Wendtland L. Molecular genetics of psychopathologies: a search for simple answers to complex problems. Behav Genet 1996; 6: Leboyer M, Bellivier F, Nosten-Bertrand M, Jouvent R, Pauls D, Mallet J. Psychiatric genetics: search for phenotypes. Trends Neurosci 1998; 1: Thaker GK, Carpenter Jr WT. Advances in schizophrenia. Nat Med 1; 7: Jablensky A. Symptoms of schizophrenia. In: Henn F, Sartorius N, Helmchen H, Lauter H (eds). Contemporary Psychiatry. Vol. 3, Springer: Berlin, 1, pp McGue M, Gottesman II. Genetic linkage in schizophrenia: perspectives from genetic epidemiology. Schizophr Bull 1989; 15: Risch N. Genetic linkage and complex diseases, with special reference to psychiatric disorders. Genet Epidemiol 199; 7: Goldberg TE, Gold JM, Greenberg R, Griffin S, Schulz SC, Pickar D et al. Contrasts between patients with affective disorders and patients with schizophrenia on a neuropsychological test battery. Am J Psychiatry 1993; 15: Finkelstein JR, Cannon TD, Gur RE, Gur RC, Moberg P. Attentional dysfunctions in neuroleptic-naïve and neuroleptic withdrawn schizophrenic patients and their siblings. J Abnorm Psychol 1997; 16: Waldo MC, Adler LE, Leonard S, Olincy A, Ross RG, Harris JG et al. Familial transmission of risk factors in the first-degree relatives of schizophrenic people. Biol Psychiatry ; 47: Hoff AL, Kremen WS. Is there a cognitive phenotype for schizophrenia: the nature and course of the disturbance in cognition? Curr Opin Psychiatry ; 15: Pardo PJ, Knesevich MA, Vogler GP, Pardo JV, Towne B, Cloninger CR et al. Genetic and state variables of neurocognitive dysfunction in schizophrenia: a twin study. Schizophr Bull ; 6: Cannon TD, Huttunen MO, Lonnqvist J, Tuulio-Henriksson A, Pirkola T, Glahn D et al. The inheritance of neuropsychological dysfunction in twins discordant for schizophrenia. Am J Hum Genet ; 67: Arolt V, Lencer R, Nolte A, Muller-Myhsok B, Purmann S, Schurmann M et al. Eye tracking dysfunction is a putative phenotypic susceptibility marker of schizophrenia and maps to a locus on chromosome 6p in families with multiple occurrence of the disease. Am J Med Genet 1996; 67: Freedman R, Coon H, Myles-Worsley M, Orr-Urtreger A, Olincy A, Davis A et al. Linkage of a neurophysiological deficit in schizophrenia to a chromosome 15 locus. Proc Natl Acad Sci USA 1997; 94: Myles-Worsley M, Coon H, McDowell J, Brenner C, Hoff M, Lind B et al. Linkage of a composite inhibitory phenotype to a chromosome q locus in eight Utah families. Am J Med Genet 1999; 88: Nelson MD, Saykin AJ, Flashman LA, Riordan HJ. Hippocampal volume reduction in schizophrenia as assessed by magnetic resonance imaging: a meta-analytic study. Arch Gen Psychiatry 1998; 55: Egan MF, Goldberg TE, Gscheidle T, Weirich M, Rawlings R, Hyde TM et al. Relative risk for cognitive impairments in siblings of patients with schizophrenia. Biol Psychiatry 1; 5: Faraone SV, Seidman LJ, Kremen WS, Pepple JR, Lyons MJ, Tsuang MT. Neuropsychological functioning among the nonpsychotic relatives of schizophrenic patients: a diagnostic efficiency analysis. J Abnorm Psychol 1995; 14: Cornblatt BA, Dworkin RH, Wolf LE, Erlenmeyer-Kimling L. Markers, developmental processes, and schizophrenia. In: Lenzensweger MF, Haugaard JJ (eds). Frontiers of Developmental 51